There was an interesting post a while back that I have just recently been contemplating, about hacking Marvel Heroic Roleplaying to emulate a 4th edition style game. While reading through it, I noticed that the author mentioned not being entirely sure that the affiliation dice worked for a D&D style game but left them in.
One of the strengths of MHRP is describing the actions that your character is taking. One of the problems with alignment is often a character just does what they do with little thought about why they are doing it, because they are in an encounter or what have you.
My thought was, why not combine alignment with affiliations for a hack like this?
How this would work is, you would pick your primary alignment at d10, one step away at d8, and one more step away at d6. This actually accomplishes two things. It gives you a "tendency," and adds some nuance to the alignment system.
So, for a Lawful Good character, you could have Lawful Good at d10, Lawful Neutral at d8, and True Neutral at d6. This is a guy that favors the law side of his alignment. Or you could have a character with Lawful Good d10, Neutral Good d8, True Neutral d6. This is a guy that wants to do the right thing, and is usually a more orderly and has rules for himself, but will worry more about doing good than upholding his own honor or codes.
Every time you take an action, you justify your alignment affiliation by explaining how that action relates to your alignment. If you are defending yourself, and you can't think of any greater purpose than surviving, you might just say, "I'm fighting to survive, and I don't want to die, so I'll take a d6 for True Neutral."
On the other hand, if you are focusing on how you need to survive this fight because you need to save the high priest that has been kidnapped, you might say, "I am keeping the high priest's safety in mind, and fighting to get closer to freeing him, so I use a d20 for Lawful good."
At the very least, having someone think about their motivation and explain what is going on in their head with each action might make alignment at least a little bit more useful as a tool for understanding and shaping the actions of the character.